Aider v0.74.0: Ollama Context & Model Support

TL;DR

• Ollama context window now dynamically adjusts to fit your chat history
• Better support for o3-mini, o1, DeepSeek V3/R1, and secondary API providers
• Multiple bugfixes: filename generation, timeout handling, .gitignore parsing, and watch limits

New

• Dynamic Ollama context window — automatically resizes to hold your current chat instead of fixed limits
• Enhanced model support — improved compatibility with o3-mini, o1-mini, o1, DeepSeek V3 & R1 via secondary API providers
• Temperature control in config — specify use_temperature: <float> in model settings for fine-grained control
• Bedrock support in Docker — full container now includes boto3 for AWS Bedrock integration
• Docker HOME persistence — containers set HOME=/app to persist ~/.aider config across runs
• R1 response cleanup — removes <think> tags from R1 responses in commit messages and weak model contexts
• Improved .gitignore handling — honors existing ignores regardless of configuration, checks .env only when present
• Watch file optimization — fully ignores top-level directories in ignore files to prevent OS watch limit hits on large subtrees like node_modules
• Faster startup — improved performance with more providers and when model metadata is cached locally
• Yes/No prompt aliases — All/Skip now work as Y/N aliases even outside group confirmations

Fixed

• Incorrect filename generation for reserved names like python, php
• --timeout flag handling
• /model command now correctly reports when weak model is unchanged
• Multi-line mode persistence through ^C at confirmation prompts

Update: pip install --upgrade aider-ai

Source: Aider